A neural harness is a system that reads neural signals, interprets them, and returns feedback — closing a loop between a nervous system and a machine.
Three verbs: reads, interprets, returns. That's the whole idea.
💡 The one-line version: a neural harness is not a mind reader. It's a translator — it listens to the body's electrical language and speaks it back in a form machines can use.
The three parts every harness shares
PART 1
🔹 Signal
The electrical activity the nervous system produces — roughly 70 microvolts per neuron.
PART 2
🔹 Interface
The physical layer that touches the body — electrodes on the scalp, in the cortex, or on the skin.
PART 3
🔹 Feedback
The return signal to the user. A buzz, a nudge, a cursor movement. This closes the loop.
🪷 The design goal: when the loop closes correctly, the user doesn't describe it as "using a robot." They describe it as "I did it."
02 · Real-time examples
Already working
Neural harnesses are not science fiction. Every example below has been demonstrated in humans or clinical trials. Each links to its primary source.
BCI2012
Cathy Hutchinson drinks coffee with her thoughts
Paralysed from the neck down after a stroke. A Utah array in her motor cortex fed a decoder. She thought about moving her arm — a robotic arm moved — she drank.
Observation mode vs Enforcement mode. The delta is the guardrail reactivity coefficient — how much a system changes when it knows it's being watched. No existing harness reports this. Jijñāsā builds it.
🧩 How to combine them. Take one course from each category. Use the Jijñāsā prompt as your capstone — a project that connects the neural signal (BCI course) to the agent loop (AI harness course) and exposes it via MCP.
06 · Applications
Where neural harnesses are implemented today
A compact reference of the fields where neural harnesses are already deployed, in trials, or in commercial products.
Field
What it does
Example
🧠 Restoring speech
Decodes attempted speech from cortical signals
UC Davis ALS implant — 99% accuracy over 3,800 hours
🖱️ Restoring movement
Controls cursor or robotic arm from motor cortex
BrainGate2 — tetraplegic patients control devices by thought
🦾 Prosthetics
EMG-controlled artificial limbs
DEKA Luke Arm — FDA-approved since 2014
🦿 Rehabilitation
Exoskeletons that trigger stepping from balance shifts
ReWalk — FDA-cleared for spinal cord injury
👁️ Sensory substitution
Camera → tongue electrode array for spatial perception
BrainPort Vision Pro — CE-marked, clinical use
✈️ Fatigue monitoring
EEG headband detects microsleep before it happens
SmartCap — mining, aviation, trucking
🎮 Consumer control
sEMG wristband reads intended finger movement
Meta Neural Band — AR/VR hands-free input
🏠 Home automation
Consumer EEG controls lights and smart devices
Research-grade BCIs integrated with smart home hubs
🛡️ AI safety
Measures behavioural change under guardrails
Guardrail reactivity coefficient — Jijñāsā Case 6
🧠 The pattern: every implementation follows the same three-part structure — signal, interface, feedback. Different hardware, different domain, same loop.
07 · The idea
What is Jijñāsā?
Jijñāsā is a single prompt. You paste it into any coding agent — Cursor, Windsurf, Claude Code, Cline, Aider. The agent becomes a build coach. You become the builder.
But it's not a fixed curriculum. Before writing a single line of code, the agent asks you five questions. Your answers define the entire harness — its domain, its signal, its output, its success criterion, and its name.
🪷 The word.Jijñāsā opens the Brahma Sutras: "athāto brahma-jijñāsā" — "now, therefore, the inquiry into reality." The method is inquiry first. Build second. Resilience throughout.
08 · What makes it unique
Six things no other prompt does
USP · 01
It is drawn out of you
Most prompts give you a fixed syllabus. Jijñāsā asks five questions first — and builds the syllabus from your answers. The harness you get is yours.
USP · 02
Every harness is genuinely unique
Two students with the same topic walk away with different files, gates, names. The ladder is structural — the content is personal.
USP · 03
Indigenous vocabulary
The harness is a jijñāsā. The build is a sopan. The resilience is a Jātaka. The vocabulary carries the tradition with it.
USP · 04
Guardrails as measurement
Twelve Jātaka precedents run in Observation or Enforcement mode. The delta between them is data — how much a student actually needs the precedent.
USP · 05
Resilience trace by the student
After every wall, the agent asks "what kept you going?" The answer is recorded. It appears in the final signature, in their own voice.
USP · 06
A signature, not a certificate
At the end, the student writes three lessons, a resilience reflection, and a one-line closing. It is theirs. No agent can write it for them.
09 · Phase 0
The five questions
01
🪷 What do you want to understand?
Not "what do you want to build" — what do you want to understand?
"How does a paralysed person move a cursor with thought?"
02
📡 What does your system read?
EEG, EMG, eye tracking, keystrokes, audio — any stream of numbers that changes over time.
EEG · EMG · eye tracking · keystrokes · audio
03
🎯 What does your system do?
Move a cursor, trigger a warning, adjust a game's difficulty, send a message.
Manas · Kriya · Drishti · Smriti · Raksha · or your own
10 · Phase 1
The bespoke ladder
Nine layers. Eight build, one prove. The structure is fixed — the content is instantiated from your charter.
L1What is intelligence in your domain?3 modules
L2Programming & your data shape3 modules
L3AI & LLMs for your signal3 modules
L4Machine learning for your task3 modules
L5Neural nets for your signal3 modules
L6Your signal — deep dive3 modules
L7The loop: read → respond3 modules
L8Wiring your harness together3 modules
L9The proof — your success criterionPhase 9
11 · The Jātaka guardrails
Twelve precedents for the wall
Every builder hits a wall. Jijñāsā holds twelve precedents from the Jātaka tradition. When the student is stuck, the agent may invoke one. Quietly. As a companion, not a teacher.
JN-01
The Long Swim
The shore is not visible. You swim anyway.
JN-02
The Patient Tortoise
Slow, steady, unbothered by the mockery of faster things.
JN-03
The Honest Word
The truth, told plainly, even when it costs.
JN-04
The Open Hand
Let go of what you built to make room for what works.
JN-05
The Steady Eye
When others panic, don't move.
JN-06
The Quiet Return
Come back to the thing that depends on you.
JN-07
The Learner's Bowl
Be the fool who asks. The asking is the wisdom.
JN-08
The Refused Throne
Walk past what's offered if it isn't yours.
JN-09
The Tender Enemy
Meet hostility without returning it.
JN-10
The Fallen Leaf
Patience under mockery. The leaf falls on its own time.
JN-11
The Waking Bird
Duty that persists through the night.
JN-12
The Silent Raft
Steadfastness in hardship. The raft does not complain.
Presence Mode
🪷 Silent holding
Agent holds the precedents quietly. Does not invoke. Used in early layers.
Invocation Mode
🪷 One precedent, when earned
After 3+ attempts at the same wall, agent may invoke one. Sparingly.
12 · Phases 9 & 10
The proof and the signature
Most harnesses stop at "it runs." Jijñāsā ends at "here is what I proved, and here is what I learned."